"Carl J. Van Arsdall" <cv*********@mvista.comWrote:
To: <py*********@python.org>
Sent: Friday, July 14, 2006 7:36 PM
Subject: Re: Python for Embedded Systems?
| Grant Edwards wrote:
| On 2006-07-14, Jack <no****@invalid.comwrote:
| >
| >Is there a Python packaging that is specifically for
| >embedded systems? ie, very small and configurable so the
| >user gets to select what modules to install?
| >>
| >For Linux-based embedded systems in particular?
| >>
| >I'm thinking of running it on the Linksys's Linux-based open
| >source router WRT54G. It has 4MB flash and 16MB RAM. I think
| >another model has 16MB flash. Any possibilities of running
| >Python on these systems?
| >>
| >
| A few years back there was a "deeply embedded python" project,
| but it's been dead for quite a while:
| >
|
http://www.tucs.fi/magazin/output.ph...0.N2.LilDeEmPy
| >
http://mail.python.org/pipermail/pyt...st/000157.html
| >
| >
| This raises a good question. Is there a need for python to change
| somewhat to work better in an embedded profile? Are there many people
| in the community interested in using python for embedded projects?
|
I for one would be interested in using it - if it can be made small enough - I
do not see it helping me with the nitty gritty of getting a system running on
bare metal, as the low level stuff in a small system is still best done in
assembler - but if there is a display and some sort of keyboard and you can
stick to character based stuff it will ease the "application" part of the work a
lot - its a natural for doing things like logging and simple transactions - and
if the hardware is strong enough to get IP running - then it will open up the
systems to be web based...
It will also be of use as you could prototype on a PC - emulators and simulators
of small systems are mostly a pain as it takes a lot of effort to describe the
actual hardware before these are useful...
"Small" in this context is an eight bit processor (like an 80x1 | x = 3, 5 )
with some tricks done in the hardware to expand the native 16 bit address space
to address say half a Meg of battery backed up static ram as data space, with
about 62k of code space available for the interpreter and the custom OS to run
in - (Harvard architecture, not von Neumann) - These processors have lately been
improved from the original about half a MIP to now give sustained performance of
better than 10 MIPS - So they may just be able to cripple along doing the
interpreting - But its *Quite a Challenge* to fit the interpreter in, though...
It is easily possible to expand these processor's address space to 24 bits -
that is 16 Meg - and it may be a better approach to make a 'Virtual Processor'
to run the python stuff - is there a definition for that sort of thing, akin to
Java's Virtual Processor? - It would be a better approach, as the resultant
higher level work would be more portable - to port to another class of
processor, you need merely write the Virtual Machine - in whatever Language
takes your fancy...
What is the minimum set of instructions and structure needed to make a
simplistic Python? - does anybody know?
Might just be a pipe dream...
- Hendrik van Rooyen
